Chunking as an object: What comes together, goes together

When people have to remember several objects, they often employ grouping or chunking. The chunking does not only enhance memory precision but also introduces systematic biases for memorized items (e.g., attraction toward the gist of the display). While demonstrated for single-feature objects (e.g., orientation, shape, or color), it remains unclear how people chunk objects with multiple features. It could involve feature-level chunking, treating each feature separately, regardless of whether chunkable features belong to the same object. Alternatively, object-level chunking may occur, where grouping features within the same object improves chunking. Previous work investigated these hypotheses and found evidence of only feature-level chunking for multi-feature objects when both orientation and color were highly similar across all objects (Song & Chong, 2023). However, this work did not put feature and object cues into direct conflict to test their relative contributions. To test this, we ran a continuous report task where participants (N=44) had to remember 6 colored orientations and later report both the color and orientation of the probed item. In one condition, half of the stimuli were similar in colors but not orientations while the other half were similar in orientations but not colors. In another condition, half of the stimuli were similar in both colors and orientations, while the other 3 had random orientations and colors. Interestingly, when both features could be chunked for the same objects (i.e., the second condition), we found greater memory precision (F(43) = 17.22, p < 0.001) and stronger attraction bias (F(43) = 9.58, p < 0.005). Our results show that chunking can occur not only at a feature level but also at an object level. This prompts questions about the interaction of chunking at different levels and the potential existence of chunking at various levels of abstraction in working memory.